Deploy and Manage Azure Compute Resources for AZ-104

Build the deployment, VM, container, and App Service judgment that AZ-104 expects from Azure administrators.

Compute on AZ-104 is broad on purpose. Microsoft wants Azure administrators who can deploy repeatably, run virtual machines competently, understand where containers fit, and manage App Service without guessing through the portal.

What this domain is really testing

The exam is less about deep software delivery and more about day-two operations. Can you read deployment structure, choose a sane VM or scale model, and manage common application hosting surfaces without breaking security, availability, or recoverability?

Current weight in the study guide

Microsoft currently weights this domain at 20–25% of AZ-104. Along with identity and governance, it is one of the most important areas to stabilize before exam day.

Work this domain in order

Start with 3.1 ARM, Bicep & Deployments, then move to 3.2 VMs, Disks & Scale Sets, 3.3 Containers & ACR, and finish with 3.4 App Service Plans, TLS & Slots.

Fast routing inside this chapter

If the scenario is really about…Go first to…
ARM templates, Bicep files, export, decompile, what-if, deployment workflows3.1 ARM, Bicep & Deployments
VM creation, sizing, disks, encryption at host, availability sets and zones, VMSS3.2 VMs, Disks & Scale Sets
Azure Container Registry, Container Instances, Container Apps, container sizing and scaling3.3 Containers & ACR
App Service plans, web apps, TLS, custom domains, backups, networking settings, deployment slots3.4 App Service Plans, TLS & Slots

Common AZ-104 traps

  • treating exported templates as perfect production IaC
  • missing the difference between scaling up and scaling out
  • forgetting that region moves are usually migration workflows, not simple property changes
  • confusing an App Service plan with the app that runs on it
  • treating every container scenario as if it were really an App Service question

If this chapter feels scattered, that is normal. The way to stabilize it is to tie each compute option back to the operating model it implies.

In this section